The Impact of the Format of Covariation Information on Causal Inferences

Representational effects in a fictitious virus-disease causal induction task were examined in three studies. In all three studies, six different judgment conditions were created by crossing two levels of virus-disease covariation (0,.5) with three levels of disease base rate (.25,.5,.75). In Study 1, the covariation information was presented as four propositions summarising the frequencies of the four patient types, namely patients with or without the virus who either did have or did not have the disease. In Study 2 the same information was presented in a 2 x 2 table with the cell frequencies represented iconistically (the presence/absence of virus/disease was shown as schematic faces that varied in expression and colour). In Study 3 the covariation information was presented in terms of a branching tree with the two main branches representing the frequencies of patients with and without the disease from which sprouted smaller branches showing the frequency of those with and without the virus. Causal judgments were poorest in Study 1, reflected significantly improved covariation discrimination in Study 2, but were most normative in Study 3. These results signal the presence of important representational effects in causal induction tasks